Weishun Zhong
weishunzhong.bsky.social
Weishun Zhong
@weishunzhong.bsky.social
Postdoctoral Member @ Institute for Advanced Study, Princeton. Recent Physics PhD @ MIT. Research in Neuroscience, Physics, and AI.
www.weishunzhong.com
Thanks for noticing our paper! I made a short 1-minute video to illustrate the basic idea — check out my post below if you're interested in learning more!

bsky.app/profile/weis...
❤️"The Physics of Romeo and Juliet": scientific evidence for "I only just met you, but I feel like I've known you for my entire life. "

📖Excited to share that our paper is selected as an Editor's Suggestion in PRL.

💡Curious about how statistical mechanics has anything to do with stories? Read on.
If you’ve seen Romeo and Juliet, you likely can summarize the plot in a single sentence that appears nowhere in the play. A new physics-inspired theory aims to explain our story-recalling ability by modeling the memory of a story as a hierarchical tree.
June 18, 2025 at 11:37 PM
I’d like to think we’ve formalized the observation of why some events can be disproportionately memorable — due to the scale-invariance of memory.

I call it “The Physics of Romeo and Juliet”: scientific evidence for “I only just met you, but I feel like I’ve known you for my entire life.”
June 18, 2025 at 11:34 PM
The finishing touch of our "story": we predict scale-invariance in memory. Our memory appear to be statistically the same no matter how long the underlying event is.

📃the paper:
journals.aps.org/prl/pdf/10.1...
June 18, 2025 at 11:20 PM
And the outcome: using this simple physics-style model, we can quantitatively reproduce how people read and recall stories—turning complex human behavior into precise and testable predictions.
June 18, 2025 at 11:20 PM
We built a minimal model that captures some key ingredients in how we remember stories - hierarchy, summarization, and abstraction. Our model is exactly solvable, and highly constrained - only two free parameters and both set to be human working memory capacity of 4.
June 18, 2025 at 11:20 PM